Port Zoning question with multiple arrays

[ColinF]

I will be zoning a couple of new B series switches soon. There will be 2 switches to support multipathing and also 2 storage arrays in the SAN while migrating from an EVA4000 to EVA6400.

Suppose the following example per switch:
port0=EVA4000 Controller A
port1=EVA4000 Controller B
port2=EVA6400 Controller A
port3=EVA6400 Controller B
port4=HOST1
etc.

My question is what would the correct way to config the zoning be? My thoughts are either:

1.Zone1=0,1,2,3,4
2.Zone1=0,1,4 Zone2=2,3,4
3.Zone1=0,4 Zone2=1,4 Zone3=2,4 Zone4=3,4

What I am unsure of is if it is correct to have storage array controllers see each other and other storage arrays? If it's fine then I suppose option 1 is correct. If it's ok for the same array to see it's other controller but not other arrays then I suppose option 2 is correct and if complete separation is the way to go then I suppose option 3 is correct.

I hope that my example is understandable, with option 1 there is 1 zone per host port, with option 2 there are 2 zones per host port and with option three there are 4 zones per host port. Right now, with our current switches I have it zoned so that controllers can't see each other. I'm pretty sure that option 3 is the correct way but I'd like to be sure. Thanks.

[Patrick Terlisten · Accepted solution]

Hello Colin,

I would prefer the third way, a beautiful, simple single HBA centric zoning. Just put one target and one initiator in one zone. This is the best zoning you can do. I would also recommend the using of WWPNs instead of port zoning. Since some years WWPN/ soft zoning is alos ASIC enforced. So it's also a "hard zoning".

Best regards,
Patrick

[Uwe Zessin]

Storage arrays do not need to 'see' each other except in replication setups.

I create separate host/array zones (e.g. host1/array1, host1/array2, ...) - this avoids 'communication leaks'. If the arrays need to talk to each other due to replication I create separate zones - This allows me to selectively disable replication traffic, which is not possible when 'leaks' exist.

I don't use fixed port addresses. Instead I use WWNs with alias names. On a Brocade switch we can use the command "nsaliasshow" and get kind of 'online-documentation':
Aliases: aSMA1_1

If we move a device, the ASICs are re-programmed and when we re-run the command it automatically adapts.

[AWilliams_1]

You can probably 'borrow' some temporary CA (Continuous Access) licences from HP so that you could use the built in replication capability of the EVAs to do your migration. It would probably be a lot quicker.
In any case I would create a zone for each host and both EVAs ie:
Zone1 = 0,1,2,3,4
Zone2 = 0,1,2,3,5
...
Then you have the full multipathing and failover capabilities available to you.